opus-2020-07-27.yml

release: eng-trk/opus-2020-07-27.zip
release-date: 2020-07-27
dataset-name: opus
modeltype: transformer
pre-processing: normalization + SentencePiece (spm32k,spm32k)
subwords:
   - source: spm32k
   - target: spm32k
subword-models:
   - source: source.spm
   - target: target.spm
source-languages:
   - eng
target-languages:
   - aze
   - bak
   - chv
   - crh
   - kaz
   - kir
   - kjh
   - kum
   - ota
   - sah
   - tat
   - tuk
   - tur
   - tyv
   - uig
   - uzb
use-target-labels:
   - >>aze_Latn<<
   - >>bak<<
   - >>chv<<
   - >>crh<<
   - >>crh_Latn<<
   - >>kaz_Cyrl<<
   - >>kaz_Latn<<
   - >>kir_Cyrl<<
   - >>kjh<<
   - >>kum<<
   - >>ota_Arab<<
   - >>ota_Latn<<
   - >>sah<<
   - >>tat<<
   - >>tat_Arab<<
   - >>tat_Latn<<
   - >>tuk<<
   - >>tuk_Latn<<
   - >>tur<<
   - >>tyv<<
   - >>uig_Arab<<
   - >>uig_Cyrl<<
   - >>uzb_Cyrl<<
   - >>uzb_Latn<<
test-data:
   - Tatoeba-test.eng-aze: 2659/10046
   - Tatoeba-test.eng-bak: 39/140
   - Tatoeba-test.eng-chv: 333/1358
   - Tatoeba-test.eng-crh: 22/81
   - Tatoeba-test.eng-kaz: 397/1668
   - Tatoeba-test.eng-kir: 118/428
   - Tatoeba-test.eng-kjh: 17/48
   - Tatoeba-test.eng-kum: 8/25
   - Tatoeba-test.eng-multi: 10000/46183
   - Tatoeba-test.eng-ota: 678/3328
   - Tatoeba-test.eng-sah: 39/131
   - Tatoeba-test.eng-tat: 1451/6996
   - Tatoeba-test.eng-tuk: 2500/12809
   - Tatoeba-test.eng-tur: 10000/49076
   - Tatoeba-test.eng-tyv: 5/19
   - Tatoeba-test.eng-uig: 3024/13084
   - Tatoeba-test.eng-uzb: 457/1514
   - newsdev2016-entr.eng-tur: 1001/14044
   - newstest2016-entr.eng-tur: 3000/44195
   - newstest2017-entr.eng-tur: 3007/45049
   - newstest2018-entr.eng-tur: 3000/45944
BLEU-scores:
   - Tatoeba-test.eng-aze: 26.0
   - Tatoeba-test.eng-bak: 9.2
   - Tatoeba-test.eng-chv: 3.9
   - Tatoeba-test.eng-crh: 7.6
   - Tatoeba-test.eng-kaz: 10.4
   - Tatoeba-test.eng-kir: 26.9
   - Tatoeba-test.eng-kjh: 2.0
   - Tatoeba-test.eng-kum: 2.7
   - Tatoeba-test.eng-multi: 18.8
   - Tatoeba-test.eng-ota: 0.4
   - Tatoeba-test.eng-sah: 0.7
   - Tatoeba-test.eng-tat: 9.6
   - Tatoeba-test.eng-tuk: 5.5
   - Tatoeba-test.eng-tur: 33.4
   - Tatoeba-test.eng-tyv: 3.6
   - Tatoeba-test.eng-uig: 0.1
   - Tatoeba-test.eng-uzb: 3.3
   - newsdev2016-entr.eng-tur: 9.5
   - newstest2016-entr.eng-tur: 8.0
   - newstest2017-entr.eng-tur: 7.8
   - newstest2018-entr.eng-tur: 8.2
chr-F-scores:
   - Tatoeba-test.eng-aze: 0.568
   - Tatoeba-test.eng-bak: 0.320
   - Tatoeba-test.eng-chv: 0.266
   - Tatoeba-test.eng-crh: 0.347
   - Tatoeba-test.eng-kaz: 0.352
   - Tatoeba-test.eng-kir: 0.508
   - Tatoeba-test.eng-kjh: 0.052
   - Tatoeba-test.eng-kum: 0.073
   - Tatoeba-test.eng-multi: 0.447
   - Tatoeba-test.eng-ota: 0.064
   - Tatoeba-test.eng-sah: 0.028
   - Tatoeba-test.eng-tat: 0.309
   - Tatoeba-test.eng-tuk: 0.309
   - Tatoeba-test.eng-tur: 0.617
   - Tatoeba-test.eng-tyv: 0.125
   - Tatoeba-test.eng-uig: 0.152
   - Tatoeba-test.eng-uzb: 0.268
   - newsdev2016-entr.eng-tur: 0.423
   - newstest2016-entr.eng-tur: 0.397
   - newstest2017-entr.eng-tur: 0.394
   - newstest2018-entr.eng-tur: 0.396